Intel announced a new tool named as LatencyTOP which can be used to find what's causing the latency. http://lkml.org/lkml/2008/1/18/355
toastyde1th is now known as toastydeath
"LatencyTOP is a Linux tool for software developers (both kernel and userspace), aimed at identifying where system latency occurs, and what kind of operation/action is causing the latency to happen. "
paranoia strikes deep... it always seems someone's posted some goody then left before i can ask
hey seb! thats a 250 uS cycle time? on the new 7i43
err, servo update time
a-l-p-h-a is now known as a-l-p-h-a
[01:29:09] <skunkworks> http://www.cnczone.com/forums/showthread.php?p=396118#post396118
he can set TLO by probing - there's sample gcode in the distribution
or... he can do a tool change in the program if he needs to touch off with a new tool
z axis touch plate would work with emc wouldnt it
custom m code?
tool length probing works great, I use it
how would one use it for a gear cutter
depends on the sensor I guess
if you probe the arbor and know the offset up to the teeth, that could work
but I doubt you'll be doing a job with a lot of tool changes, one of them being a gear cutter
well one needs to rotate the spindle to know the peak
I picture tool length probing as more useful for (say) loading a series of drills in a jacobs chuck
been contemplating optical of some method
can be sensitive enough to work as a touch pad
[01:48:21] <Ziegler> http://en.wikipedia.org/wiki/Piezoelectricity
combine with a simple transistor and the probe pin on the para port
hmm possible, I actually want 3 measurements for automated centering and offset for the cutter
and some of the cutters are damn small
or a line ccd camera(scanner optics maybe)
I just got about 10 lbs of machinable wax
current methods can take a day to make 1 pinion
been trying to make some molds for it to reform waste into "bricks"
day to make a pinion?
what are your current methods?
(im just curious)
microscope and trial and error
no 1950's production gear cutter
got any pictures?
once set up one can make 100's , yes link in a sec
www.archivist.info/cnc top right
now thats a newat machine
its massivly built for the size it cuts
mass is good (as long as you don't have to move it down steps)
down is easy up is harder
hmm a jeweled v block could sense od and centering with x/y motion detection or combination of v and flat
you mean the contacting surfaces would be jewels, or the block would pivot or something on jewels?
with directionam sensing (peizo or similar)
doesn't that risk dulling the cutter?
a modified record pickup
well if slow enough
needs to be hard to avoid wear and be repeatable
* jmkasunich imagines something like those old pulsar tuning fork watches - a vibrating probe - when it touches something, the vibrations stop
if you had a probe (as opposed to a v-block), how small would it need to be? 1mm sphere? 0.5mm sphere?
cutter tooth width is 10 thou on small sizes
do you need to sneak in between teeth?
just looking for peak od and centerline
this is like a traditional gear cutter, only smaller, right? multiple cutter teeth, but a single gear tooth - not a hob
cut a pinion and see how good and mod offsets and....
you say you use a microscope.... would optical methods work (webcam looking in microscope eyepiece) or is it something else that makes it a try-and-repeat process?
form of cutter is often not exactly simetrical
can you 'look for daylight' ?
I need to do some pics of the problems I see
when you say not symmetrical. do you mean tooth-to-tooth variation on the cutter, and you have to find the biggest tooth?
(actually, it might be one tooth cuts farthest left, another cuts farthest down, and a third cuts farthest right
or swap the tool for something that is symmetric while finding location, then mount the real tool for cutting
sum of all errors cutter, arbor, spindle
suppose you can achieve a bright white background, and the cutter is dark
video it while running slowly, then merge all the frames, using a "darkest pixel wins" algorithm
the final dark outline shows what will be cut
I do have some backlight units to try that
doing that I could correct for form to some extent
i'm hearing hints for video recognition in hal/emc :)
would be fun
tomp: not really
this would be more of a touch-off type step
but instead of using an edge finder, you'd be using video
'looking for daylight' video
'lookin for daylight' is watching for when something approaches soemthing by watching when light is no longer visible between them
and a pile of maths to get best depth depending on cutter profile and module(DP)
gearcutting is always a pile of maths
as ours is cycloidal
or near enough
looking right being more important than actual shape
lewin1 is now known as lewing
can someone look at the last row on page 3 of measure.igpp.ucla.edu/hemisphere/Manuals/ads7870.pdf and tell me what "aperture delay" means?
pew pew pew
arghhh this a/d takes 577 cycles of a 29.49MHz clock just to convert a single channel
obviously not intended for high speed work
pick a different one
(assuming you need high speed)
is it the norm among a/d chips?
there is no such thing as "norm"
there are ADs ranging from 60 samples per second to 60 million (or more)
and from 8 bits resolution to 22 bits
and from $1 to $100
one benjamin for just one chip?
high resolution at high speed = big bucks
what would they be used for?
digitizing rader waveforms and stuff like that
argh the doc claims a 52kHz conversion rate
thats about what I calculated
577 clocks at 29MHz = something like 51 KHz
what rate do you need?
how many bits of precision do you need?
how would I know?
how do you want to interface to it - serial or parallel
wtf are you looking at A/D converters for if you have no clue what you are gonna use it for?
er, measuring voltage? 120VAC scaled down to 0-20V?
I'm just not sure how many bits I need
well, why are you measuring it?
so you can read it with your eyes? then a few samples per second will do
how precisely do you need to measure it?
+/- 1V? +/- 0.1V?
i see what you mean
what about interface? isn't serial usually more troublesome than parallel?
it depends on what you are hooking it to
if you are using a microcontroller that has SPI ports built in, serial can be more convenient, uses fewer uP pins, and is usually cheaper (smaller package because of fewer data pins)
wow - how about $978 for an A/D converter
National Semiconductor, ADC083000 - 8 bits, 3 gigasamples per second
isn't SPI just some underlying protocol? what do you mean by built-in spi port?
what quantity for the 978?
thats the 1 piece price
if you buy 1000 of them they're only $808 each
cool... lets all ask for samples
(digikey pricing - except nobody in their right mind would buy 1000 from digikey - if you are ordering 800K of chips, you go direct to the manufacturer
isn't digikey aware of that?
somebody might buy 1 or 10 or even 100 from them
it doesn't cost them anything to list a 1000 piece price on the website
so, your A/D requirement
do you have any clue what sample rate and accuracy you need?
why are you measuring the line voltage? for fun? to control something? to monitor something?
...the lights in my dad's office flicker
ok, so you really do want to monitor a 120V 60Hz line
he thinks the landlord is cutting corners with the electrical wiring, but he has no proof
I assume you don't have an oscillscope, or you'd be using that already
scopes can take 20V?
with the right probes
* maddash goes check
usually the scope can take 5-10V at the connector, and with 10x probes that 50-100V at the probe tip
if you don't know EXACTLY what you are doing, you better not try to measure line voltage
yow! the wires got so hot!
don't tell me you just tried to measure line voltage with your scope
something must have happened to the scaling circuitry
you probably just killed your scope
wtf were you thinking?
no, no I didn't hook up the scope yet
I'm not *that* stupid
what wires got hot?
er, s/wires/rectifier chip/
or whatever is the name of the chip that converts ac to dc
* jmkasunich is very confused
you have some circuit that takes the 120V ac, reduces the voltage, and turns it to DC?
and you want to monitor that DC with an A/D converter, or a scope, or whatever?
yes to both
lm2575 = boost regulator, 6 volts MAX input voltage
how can that possibly be part of a 120V to 20V circuit?
that's odd, because it's the hottest chip on the board
well, how many volts are you giving it?
crap, now wonder it's not working
I forgot that I'm using a power brick
why are you messing with chips like that anyway?
what else would I work with?
since you can't seem to tell me what you are trying to do, I don't know
i am impressed by you guys
but switching power supply chips are not for beginners
all I know so far is flickering light,. bitching about a 50KHz A/D being too slow, and a boost regulator that is intended to take battery power and step it up to 5V or so
how's it 5V?
what, the LM2757?
brb, I think I shocked myself
ok, my dyslexia got the best of me
2575 is a step down switcher
this is stupid - I have better things to do than try to figure out riddles
* jmkasunich mills dovetails
* maddash reaches for the PETA hotline number
its OK, I only use the tails off of organic doves
none of this factory farm dove stuff for me
free range doves!
what are you dovetailing
a couple blocks of cast iron that will clamp onto the bedways of my shoptask
they're t-slotted on top, so I can attach things like a steady rest
but can they run linux?
the blocks of cast iron? I doubt it
is there a bug list for emc? I'm just wondering where jepler got "bug #1735431" for shcom.cc
its on sourceforge
[05:18:23] <jmkasunich> http://sourceforge.net/tracker/?group_id=6744
haha, cradek wants donations
er, higher priority numbers denote greater urgency, right?
or the reverse?
high = high
am I allowed to submit fixes for bugs that are already assigned to someone else?
sweet. I'm going to be famous.
i just wrapped the signal wires for my encoder in Aluminum foil and now they dont glitch all over the place when I strain the motors :-)
servo people: can you recommend wires & connectors for motors & encoders?
other than "shielded twisted pair" (for the encoders) I don't have any specific recommendations
twisted pair even for non-differential encoders? run ground on the other wire of each pair?
fewer connectors is good (typical failure point, so try to have least interruptions in any signal)
right, mechanical failure and electrical discontinuity
on the other hand, connectors makes the system nicely modular, especially useful when you're prototyping & just playing around
wasnt the 1st semiconductor a bad joint ;-)
heh, i donno
wow, some of these bugs are really OLD
am I to understand that emc2 has no more than 17 open bugs at this time?
Hey folks, sorry to jump right in, but I'm hoping someone can provide some info and opinions...
First, anyone worked with the electronics from probotix.com?
Second, does EMC have an equivalent to the automatic tool zeroing that Mach3 has?
emc has probing and coordinate offsets..
not sure if you can set tool lengths from within g-code, but surely it's possible with a script
is probing a general function or something specific to a single task?
probing means emc moves in some direction until a switch is closed (or opened)
look at g38 and g41
er.. not g41, g43
axis has a 'touch off' button but i haven't used it so i dont know what it actually does
I was just reading about it in the wiki
seems similar, but he uses a feeler gauge and manual jogging to 0 position
g38.2 z-#<probe depth>; g43.1 K #5063 should do what you want
; means new line
i wonder why probotix uses screw terminals for their kits
you dont have to use them it seems, if you use all their electronics there is a ribbon cable to go between the boards
is that what you were asking?
so the terminals are just along for the ride?
[06:14:25] <brendanjerwin> http://www.probotix.com/accessories/
10PIN 4" IDC Cable
I think the terminals are to allow for extensibility.
use some other breakout board if you want... that kind of thing
brendanjerwin: the 'touch off' does g10
fenn: what is the 5063 in your example? the offset of the tool set by the probe?
ok, how did you know what value would be placed in 5063?
i guessed :)
[06:18:33] <fenn> http://www.linuxcnc.org/docs/html/gcode_main.html#sub:G43,-G49:-Tool
5061 is probably X, so 5063 is probably Z
wow, you really did guess.... I thought that was a bit of a RTFM coming from you! :)
oh, -#<var> won't work
you'd have to do some crap like [0-#<var>]
I've just come to the realization that my concept that I'd be going quickly from CAD to milling without hand adjusting the g-code just isn't going to happen... :) I need to get comfortable with this stuff I guess.
maddash: how did you know touch off does g10?
i recalled peeking inside axis's code half a year ago
brendanjerwin: if you have a "real" CAM program there are post processor templates where you can do weird stuff to get the g-code output you want
but you have to know what you want first of all
yep, not a problem, just a puzzle. I have a lot of learning to do.
is it common for people to use SCM packages (CVS, SVN) for g-code stuff? Is a g-code program something that is commonly tweaked over time, like a 'regular' program?
not in my limited experience
the machinists I work with don't save prior versions at all, and usually just re-write the whole thing
not kidding at all?
wow, seems painful...
to have to re-write each time...
most handwritten code isn't long
and if there are changes they usually demand a very large change
so the amount of work to revise vs. rewrite is similar
it's not at all like normal coding
ah... I see.
hence the post-processing stuff on the CAM packages
so that machine specific stuff can be injected
speaking of CAM; I was thinking of focusing on using CamBam. Any thoughts?
this is one cool puma
[08:58:59] <maddash> http://gizmodo.com/346294/70000-industrial-robot-modded-to-fire-flameballs
even though those guys are idiots
way cool: http://www.wpyr.com/multimedia/images/GH01.jpg
hello mr Joni
how's your crap going?
the crap is doing poorly
to many school projects thats eating time :/
* Vq^ disappears
On the Stepconf Basic machine page their is Step Time and Step Space. The specs for my 203v list step pulse "0" time and step pulse "1" time
which one goes where?
[15:38:00] <Ziegler> http://producten.hema.nl/
fsdafsd is now known as Unit41
Just hit a small speed bump: yesterday my system would load emc and today it complains
core_servo_hal:20 pin 'axis.2.motor-pos-fb' does not exist
AXES = 2
If I change to AXES = 1 then it drops to axis.1....
core_servo file is the unaltered dist 2.2.2-1 motenc
got to go now, back later, sry
the only thing that has changed since yesterday is the temp of the shop .. probably down to 60. ... 23 degrees outside.
dave_1: if it worked yesterday but not today it is a HARDWARE problem. do not monkey with your working software configuration
I wonder if it isn't a cold disk ... and it just shows up this way.
really want to get this thing up on one axis so I can try at_pid
is anyone commenting on at_pid?
the real error is probably in dmesg. could be "I can't find the motenc card"
well at least I can check that ... but the card hasn't moved since yesterday
I'll wander down and check dmesg
cradek loads all the rt stuff just fine ... then starts on io and shutdown
dmesg doesn't give a hint
maybe pastebin the stdout and dmesg for others to see? maybe something is hiding in there
I'll work on that
bbl ... again
oh, yes ... how do I clear dmesg so I can start over
sudo dmesg -c
dave_1: I have played with the at_pid. It looks like it would work great but my setup (just a servo) didn't have enough mass to do a zn tuning calculation (period was to low) petev has used it (as he wrote it)
I left out the sudo .. tnx
it got me close
close is good ... then one can tweak
gotta go see if I can get a hint on my system glitch
this new biostar p4 motherboard performs almost exactly the same as the old ecs motherboard except for memory write speed (ddr vs ddr2) and disk speed. (almost twice as fast)
(just swapping processor)
seems zippy :)
makes sense I guess.. the memory in the old computer is 333mhz vs 667mhz
cradek, etc. I cat'd emc2 errors and dmesg
[18:40:45] <dave_1> http://pastebin.ca/864286
I don't see much useful but must be skipping something important
cradek most interesting ... same failure mode WITHOUT the motenc card plugged in!
I think I ate too much kim-chee last night
ah ... tums time
SWPLinux: Uh oh... I was wondering why I heard the Haz Mat tem being dispatched.
maybe the bland pancakes for breakfast will help :)
soak up some of the garlic and acid???
plus the added acid of the coffee (required)
caffeine is good stuff
maybe I'll make more
pancakes or coffee?
coffee - don't have full cooking facilities in this hotel :)
ah .. so you are not home
I'm in your time zone
(Santa Clara actually, but who's counting)
Only 800 mi south
a mere stones throw
(compared to usual)
if you fly
nasty days drive
or have a very god stone throwing arm
a day to SF .. a day to SLC 2.5 hard days to Galesburg
SWPLinux: There was this place that you can pay by the hour to use things like a mill, laser cutter, plasma cutter, etc, but I swear I can't remember the name nor find it in my bookmarks.
hmmm. sounds like a good thing to try to remember :)
in this area?
I've got an interesting problem .. emc is shutting down ... on trying to load io ... shuts down before checking for card.
Jymm check on ebay
[19:15:50] <SWPLinux> http://search.ebay.com/Machine-Shop-Time-CNC-Lathe-Mill-1-In-Quality-Work_W0QQfsiZ1QQsatitleZMachineQ20ShopQ20TimeQ20CNCQ20LatheQ20MillQ20Q2dQ20Q231Q20InQ20Q51ualityQ20WorkQ21Q21
Yeah, Not in Santa Clara county, but someplace reasonably close.
SOB I FOUND IT!!! http://techshop.ws/index.html
hey dave - got a ballscrew holding question, I wonder if you can give some insight into what might be the best course of action
so, this is Jymm's machine
the X ballscrew is held in place by a single bearing at each end
you tighten nuts on both ends of the screw to "bind" it to the frame
(the gantry moves with the ballnut)
so, it seems to me that capturing one end "hard" - between nut and a shoulder on the screw, would be better
I should draw a picture of how it is now, it's hard to describe
[19:22:06] <Jymm> http://farm2.static.flickr.com/1213/863197056_71deace0d0_b.jpg
(part of why Jymm was having a hard time :) )
In that photo, the gantry has been disconnected from the ballnut.
did you put it back in, or is that the old photo you had?
so, if you look at the two mounts on the sides, they have a flange on the inner side, and a nut on the end, but I think neither one actually clamps the inner race of the ball bearing from both sides
That Al 'U' that the ball screw is mounted in does't look too stiff.
so the question is, just grab one nut hard and let the other float a little (like a BP)
or machine a new mount for one end and capture a pair of angular-contact bearings?
bearing are single taper roller?
heh - nope
6000 skate bearings ...
dave_1: The U channel is more of a "dust shield" than anything else.
let me look at some stuff brg
it seems to me that the current setup is basically using the ballscrew and two nuts like a bearing puller on the inner races :)
they should handle that to some extent
as long as not over tightened
well, the problem is that the only thing that fixes the screw to the frame is how tightly you cinch those nuts down
SWPLinux: Since we were tlaking about it last night, I do remember measuring the threaded holes in the 80/20 and they were drilled/tapped correctly - no skew in them iirc.
ok, that's a good thing
4 small bolts to frame they are separate (what is looks like on pic)
dave_1: you still around?
making an adjustable holder could be difficult without a mill :)
SWPLinux: And when you move the underside gantry plate in place to attach the ballnut mount, they're pretty much aligned.
regarding your "core_servo.hal:20: pin 'axis.2.motor-pos-fb' does not exist" error message
I would be checking squareness in case of screw bending/binding
I'm about 99.94% sure that core_servo.hal tries to set up three axes
if you configure EMC for 1 or 2 axes, then the axis.2.whatever pins will be missing
archivist: oh, it's almost definitely bending/binding, it's just a question of how many things need to be changed to make it work
1 axis = axis.0.whatever
2 axes = axis.0. whatever and axis.1.whatever
jmkasunich: we should have for loops in halcmd :)
for i=0..num_axes / connect thing.i.whatever / next :)
so comment out axes in core_ ?
yeah, assuming you only want 1 or two axes
or just tell emc that there are three, and don't use the third one
I suppose I can lie to emc ... it doesn't know any better
depends on what the permanent setup is gonna be - if its a three axis machine and you are using two for testing, I'd tell emc its three
expecially for just getting set up
if the machine actually only has two axes, I'd fix the hal files to match
I'm going for one first ... then two and maybe three eventually
back to Jymm and SWP
mount ball screw ends in angular ballbearings or better yet high angle ball bearings.
secure with nut and spring washer so it is always tight.
so bearings tight to screw, but not necessarily to mount, like the "far end" of a BP?
I don't know what proper end loading is. ... tension on the ball screw
but on one side, captured tightly on the frame and screw
I suspect tight to mount and screw ... at least on one end
similar to the BP arrangement
this is steel and aluminum with a ~40C temperature span, so thermal expansion could be an issue (like the BP)
good axial and radial support on one end, and radial only at the other end (my 2 cents)
_especially_ since there is a mixed metals situation
ok, good to have as many cents as possible :)
gotta run ... too much going on today.. see ya later
see you - thanks
I did something very unusual today
if you try to pull on the screw from both ends, when it heats up and expands you lose the tension
I went to HGR surplus and came home emptyhanded
uh - should aI ask what? :)
oh... my... god!
we've got to get this man a hangar
fenn: good point
he has one, it's called HGR
if I had a hanger, I'd be in deep serious trouble
well of course - HGR is open the third saturday of every month?
with the current situation, I usually can only buy small stuff
(which is the later Saturday of CNC workshop)
SWPadnos: yes - they call it the saturday sidewalk sale
Jymm and I went to a couple of surplus places yesterday - one of them had a Van Norman mill
which model? #12?
dunno - I didn't look too closely
how big was it?
it's not the same as yours, but I think it was a horizontal/vertical
hmmm - taller than me, so 6' +
smaller than a bport? kind of short and squatty?
no, relatively tall
thats probably one of the big ones then
I think. there were two large-ish machines there and I may be melding hem together
they had semiconductor fab stuff too, like a 4-station spin-coater
the #12 is a nice home shop size, the others are beasts, of varying amounts of beastishness
heh - this was probably a samaller beast then
the top of the motor is probalby at top-of-head height for a 6' person
maybe even eyeball height
hmmm. I don't think it looked like that, but I could be wrong
I also didn't notice if it was on a pallet or something, so it could have been shorter than it looked
the common feature of all the vn ram types is the swiveling cutterhead on the left side of a sliding ram
I didn't look too closely, as I'm going home in an airplane ...
[19:48:35] <jmkasunich> http://jmkasunich.com/vannorman/VN_12_Horiz.jpg
[19:48:39] <jmkasunich> http://jmkasunich.com/vannorman/VN_12_Vert.jpg
head in both orientations
I just saw one at HGR today
hey Jymm - the TechShop is open today ...
how does techshop handle insurance/liabilities? letting people off the street at a cnc mill, welder, ripsaw ! sounds like a great service, one i might use.
shame there isn't one around here
heh - that's a business I would love to franchise
there's times I'd like to be able to weld stuff, but I don't want to get a complete welding setup
jmk, start one with equipment from hgr, offer to sell them ( dont buy them :-)
"them" = what? welders?
them = the equipment, I think
all the tools TechShop has, from HGR
I wouldn't touch running a techshop with a 20 foot pole
SWPLinux: looking for bearings, not even McMaster has them - smallest ID is 12mm
too much paperwork, insurance, etc
Jymm: what kind of bearings are you looking for?
Jymm: yeah - I'd like to try just clamping one of the bearings hard to see what that does, but later maybe make blocks with a larger bore
jmkasunich: angular contact 10x26x8 (or dbl being 16mm)
jmkasunich: angular contact version of 6000
arnt skate bearing id pf 7mm?
10mm ID, 28mm OD, 8mm thick?
they're mostly 10 ID x 30 OD
the ones he has now are 10ID 26OD
skate bearings: od=20mm id=8mm thick=7mm
still too big?
these are 6000-2RS
skate bearings don't matter
not big enough :)
skate != thrust bearing
what matters is the size of his ballscrew
2nd thing that matters is the bore in his housing - although a new housing could be made
I didnt know you were looking for thrust bearings
third thing that matters is thickness, within some range
I think that'll be needed at some point
(a new bearing block)
real men pour babbit
mcmaster has 10x30x9 in angular contact, for $23 each
yep - that would require boring out the block - it's only 26mm at the moment
how long is the turned portion of the screw? if there is only one bearing per end now, its gonna be tough to find two that will fit on one end
Koyo lists 10 x 30 angular contact 10 x 30 x 14.3 #3200
I think there's enough room
18mm long bearing seat?
is the 3200 series an angular contact pair?
JMK going back to 3 axes fixed the problem
do they have a long spacer on there now or something?
jmkasunich: the screw almost floats right now
now I can get on with life
SWPLinux: not following you
it's only captured because there's a nut on both ends, pushing gains the bearings
right, that I get
but the ends of the screw have some machining
Jymm: can you measure the section of the screw that's 10mm diameter?
a shoulder (even if its just the start of the ballscrew part), then a smooth section, then a threaded section for the nut
i assume the machined end is not long enough to fit two bearings along it
fenn: I suspect the same, but I try to avoid assuming
yep - the shoulder is the end of the threaded part, then there's a fairly long smooth section before the thread for the nut
SWPLinux: In ebonics.... "we be fscked"
no - go measure :)
Jymm: that isn't a measurement
this is a router, right?
I think it was fairly long, possibly to account for screwups in positioning of the mounts
so we're not talking machine tool tolerances here
well, "closer" would be better :)
hi juanjo. have you made progress on your machining center?
On the end where the pulley would mount... From the should3er to the first thread .250"
Jymm doesn't have a lathe does he?
hello !!!, well i move it to a suitable location to be retrofitted
Jymm: what about the other end?
Jymm: so 10mm diameter for 0.250 long, then threads?
Other end from shoulder to first thread is .350"
ok, I'd use that end I guess :)
this was a very hard job, due the weight ans size of the machine, i need a 20 ton crane and 2 flat truck
to move it
Let me grab camera...
at the pulley end, how is the pulley attached
yes, do that ;-)
is the 'angular contact bearing' vs a thrust bearing to try get some radial and thrust qualities in a single package? (.250 is too small for 2 bearings, ditto for .350)
I'm looking at the flikr photo now, but it doesn't show much detail
tomp: the angular contact is to get thrust, yes
I'm thinking about the possiblities of needle thrust bearings
they make some thing ones
bearing plus two hardened washers
needle bearing + angular contact?
no, needle bearings (two) for thrust, and whatever for radial
if the bearing blocks are only designed for one bearing, it's going to take some lathe work anyway, so you might as well make an extension
at the belt end a cheapo ball bearing would be nice, and the other end it doesn't much matter, could even be sleeve
since there is little or no radial load at that end
the move was a week ago, the temperature in tucuman was 38 degreed celcius, so is like you working inside a Sauna!!!. in this week i study the gantry type robot tool changer, and the way to replace the servos for the axis x1,y1,v by a Bosh type SD
mcmaster carr 5909K11 10mm needle
a needle thrust bearing is 2mm thick, and the hardened washers are 1mm each
so you can do a nice thrust bearing in 4mm
hmmm - cool
and tbey're cheap
the bearing is $3.05, and the washers $1.03 each
I'm using those on my shoptask Y axis (have been for years)
Jymm has screw in an aluminum frame, gotta be careful with tension
uh, so 8mm for two needle bearings leaves .036" for the radial bearing
fenn: no, you gotta be creative
which is why we're waiting for Jymm to come up with pics
yeah but if you're being creative, why bother messing around
for example - put one thrust bearing between the bracket and the pulley
(only works if the pulley can handle thrust, which depends on how its attached to the shaft, which depends on us seeing a picture
2, I think
also i very confusing abouth a realible i/o i see the vital system DSPMC/IP interface, anybody kown if it i can use it with emc ??
the pic shows ballscrew threads, then bearing mount block, then pulley
so where is the nut that is clamping the bearing? inside the bearing mount block?
(this is the right end in the pic)
jmkasunich: yes, I think so
so the screw is smooth, then threaded, then smooth again for the pulley?
juanjo: no, emc does not support the DSPMC
and I think a smaller diameter, like 6mm for the pulley
has to be smaller, unless you have a magic nut
(I think they didn't want to bore out the pulleys, so they made the shaft match)
I need to wait for the photo - I don't remember the specifics :)
how is the bearing retained in the block? its bored from one side, with a shoulder at the bottom?
it was two days ago that I looked at it
right - it's not fully captured, it just has a shoulder on the screw side
so even if you could fit two angular contact in there, you'd need to add something to clamp the bearing into the block
why the recommendation for two needle bearings? Are those supposed to be at the same end of the screw or opposite ends?
a needle thrust bearing only supports thrust one way
odd - it looks symmetric
you'd do something like one on each end fof the bracket
it is symmetric
but if one washer is attached to the bracket, and one attached to the screw, it can only support force that pushes the washers towards each other
ah - ok
ideally it would be like this:
it's quite obvious when you pick one up and it falls apart
right - you always have to push, it doesn't help if you're pulling
screw shoulder - rotating bearing washer - bearing needles - stationary washer - mounting block - stationary washer - needles - rotating washer - double nuts, locked to each other
and you might be able to fit a radial bearing inside the mounting block between the stationary washers
ok humm this was a good option to my proyect due de stand alone plc,
double nuts so they don't come loose, or for another reason?
so they don't come loose
one nut and loctite? :)
sure, if that floats your boat
i want to put a sort of supervisory board in the event of PC hangup to take the control to a safe stop
the issues I see with jymm's specific case are: if there isn't a real shoulder, only the thread, then the first rotating washer won't sit flat
the shoulder has to be daed square, not sure it is `( else needle bearing is wobbly)
juanjo: you can use an external charge pump circuit in the e-stop chain
yes i designing one with a AVR and triger it at a interval by a pin of the emc.
you should use hardware, not software for that
even AVR software :)
ok by example a 555 timer
sure, or a diode/capacitor + relay/transistor "switch"
juanjo: usually charge pumps are very simple
two caps, two diodes, and a transistor driving a relay, for example
Of course the battery is dead, silly me - charging now
have sense, we not be be hold on soft
anybody have a idea the scan time spected for every pin, using a 1ghz machine???
with what hardware?
with motenc-100 by example
don't know about that hardware :)
1ms is the default, and is quite easy
do you need faster?
the speed of the CPU isn't all that important for I/O
100us should be possible i think, but it depends what is going on
it also depends on the hardware
not i need to get a stimeted to see the times in the cicles for the machine, due the large i/o used , 250 imputs
at 1ms i get 1/4 sec to scan the whole boards
the inputs are read in parallel
when I say 1ms, I mean that EMC reads all the inputs every 1mS
the driver should read all the inputs every cycle
but you don't get anywhere near 250 I/Os with any single board I know of
I don't know what hardware you are using to handle 250 inputs, since the motenc only has 24 or something like that
ahh ok, i comfusing the question
unless there's a modbus thing in there :)
* fenn is curious what 250 inputs are used for
i thinking in a secuentia scaning, not parallell
the default is parallel, although a form of sequential might be used to expand the I/O count
ugh.. with multiplexers or something?
modbus, for example
yes, or shift registers, or whatever
spi over parport to a bank of avr's :)
SWPadnos: what I'm thinking of would be RT
so 1ms is fine to this
sure - SPI clocked by the driver
(more or less)
I was thinking more of a bank of $0.25 shift registers
jmkasunich: how do you clock them 8 times in one i/o cycle?
a 24 I/O board with say 20 x 8 bit shift registers and one i/o for clock would give you 160 inputs scanned every 8ms
fenn: I wouldn't
Ok, shitty pics, can retake when other camera is charged, in the meantime...
250 impust, this machine have a lot of sensors like the 6 doors, the 4 panels with buttons and selectors, have a full line of sensors for each of the 7 axis, the 2 pallets
alpha1125 is now known as a-l-p-h-a
pulley end http://farm3.static.flickr.com/2161/2204653656_f0f27ff6e9_b.jpg
get some focus, will you? ;)
SWPLinux: Ok, shitty pics, can retake when other camera is charged, in the meantime...
ok spi sound good also the shift registers!!!!!!
damn. I didn't remember it being that stupid :)
fwiw the hal_m5i20 driver seems to take about 25us to read and write everything (dac, digital I/O, encoder) when run from a 1ms thread on a 2.1GHz machine. digital-out-write takes less than 1us, while digital-in-read takes less than 3us. the m5i20 is a PCI card. more info: http://pastebin.ca/864691
heh - phone camera?
SWPLinux: what was the problem that lead to thi discussion anyway?
well, one of the bearings had gone oval
lash? or things loosening up?
floating (?) end http://farm3.static.flickr.com/2023/2204653568_a97cf23308_b.jpg
wobble wobble wobble
you could feel it when pushing the gantry by hand
i have a rack with a PLC i/o boards the Siemens 130W, and i figuring to crack it with a aver to convert to spi
but the screw is fine when it's not mounted
so the actual bearing is shot?
(checked by Thompson engineers as well as me)
one of them is, but he's got replacements
of the same kind - radial
since the mounting method is the equivalent of a bearing puller though, it seemed to me that just putting in another bearing wouldn't be the best thing in the long run
if they are deep groove radial they can probably handle non-trivial axial loads, at the expense of a few thou of lash
what if you locked one bearing to one end of the screw, and locked that bearings outer ring into the bracket
I think they're cheap groove radial
lock the other bearing to bracket or screw but not both
due the board is made to a round trip to the moon!!!
that was my idea for the first step in figuring this out
using the bearings he's already got, and the mounts as well (plus a capture cover on one of them)
jepler: stepconf is very nice, you might consider adding the gecko 203v to the stepper list it has much different timing than the 201 and 202 thanks
stepper driver list
SWPLinux: mcmaster has 6000 series bearings listed under "electric motor quality" - I bet those are deep groove
they could be
wobble wobble wobble (not a turkey) maybe bearing not perp to axle?
when a motor is mounted with the shaft vertical, the bearings have to handle the weight of the rotor as an axial load
6000 series are often used in 24/7 motors
what kind of forces are expected on this machine? 10s of lbs, 100s? not 1000 lbs
they wont mind the load
6000 is rated ag 1020 lbs radial
tom1: I could feel the wobbling when I was moving the screw in my hand - as well as on the machine
I'm sure they can handle 10% of that as axial, probably 25%
ok, that's probably enough
I think the easiest thing to do is to make a plate to hold a bearing down hard
in the mount
those wonderfully fuzzy photos don't show the shoulder very well
if the shoulder is really just screw, the bearing may not be seating square on it
next best thing (lots harder) is to make new bearing mounts and a new "foot plate" for the nut
woah mcmaster searches as you type
that could be
the inner races have a chamfer, and if the minor diameter of the screw sits in the chamfer, but the screw thread itself contacts the face, the bearing will be tilted
hmmm. that could be part of the problem - I could only feel the wobbling in one direction - when pushing (I think)
this sounds like a job for a lathe...
629 mearings from mcmaster hare the same OD and thickness, but 9mm ID - any chance of remachining the screw down to 9mm to leave a better shoulder?
though I could also tilt the inner race by hand, so that bearing was definitely shot
BigJohnT: argh, I knew the timing on the 203v was different, I didn't know they were incompatible. 201: .5uS/4uS 203v: 1uS/2uS (step "0" and "1" times)
machining the screw would probably be the hardest thing to do at this point
does jymm even have precision tools? like a mic?
jepler: can you just take the max of each?
it's likely the hardest thing on the machine, and neither of us is a lathe master
for all we know the shaft is 9.7mm instead of 9.98 mm
jepler: yes, I'd make the change but I can't figure out all your code LOL
oh, the bearings slide on and off quite easily
er, wait -- 203v is 2uS/1uS which is even worse
SWPLinux: wrong answer
also, they can be "press-fit" into/out of the blocks by hand
guys i put a image of the machine in
inner races are usually supposed to be press fit
[20:49:52] <juanjo> http://imagebin.org/13218
the direction is much faster on the 203v
yeah - press fit with a tool, not a finger
juanjo: nice machine :)
and outer races slip fit
who made this machine anyway?
I usually thing of them as "tap-fit" - with a soft hammer :)
BigJohnT: but having direction too large doesn't hurt much (doesn't lower max step rate)
ok didn't know, I put the values in manually but thought it would help others
a fleshy hammer
rubber/plastic - the same one I use to tap the drawbar
juanjo: real nice, twin palette
tricky part is applying force only to the race
yep - I use a tube for that
anyway, if the shaft is loose in the inner race, IMO the shaft is fscked
Yel Tom1 , is a big iron
needs to be measured with a proper mic to see how fscked
juanjo: perhaps you should start out with something smaller and easier, just to learn emc
yep. I didn't bring my measurement tools, though Jymm has a dial caliper at least
and everything is big the mains fuses are 100A
what did the caliper say the bearing seat diameter is?
jepler: just wondering does stepconf use another file for the GUI part? or do I need to study it harder?
we didn't measure it while I was there
Jymm: got those calipers out?
calipers aren't good for much better than a thou or two anyway, unless you are very careful
BigJohnT: yes, the gui file is "stepconf.glade". that's where the names of the different drive types are mentioned
yes fenn i have year to made tha conversion so i thr the emc on a desktop
turning that to 9mm would also help with seating the bearing race to the screw
right, assuming it didn't wind up undersize again
the other problem with the 9mm plan would be if the nut thread is 10mm
indeed. it was a little smaller I think
unfortunately, it looks like I've written it in a way that makes it disruptive to add new drive types, because I use numbered indexes. adding something either means making "other" an item in the middle of the list, or making stepconf files forward-incompatible, or making more extensive changes to use names instead of indexes
i want to buy the hardware to support it and asembly it on the bench to figure how solve it
loctite makes a bearing retaining compound intended to fill gaps up to a couple thou between race and shaft
yep - 682 or something
the trick is getting it on centered and perpendicular
jepler: you can hack it so other is always the end
I bought some, in case I screwed up boring out my pulleys :)
I guess the bearing block could hold it perpendicular, but centered is tougher
i see the time in the m5i20 nad very impresed with it
if the bearing block isn't a POS
not having a lathe is crippling for this
I've been crippled all my life then :)
fenn: those would be the "more extensive changes"
if I was gonna loctite it, I'd bore a piece of aluminum bar with a good slip fit over the screw, and a pocket for the bearing
slip it on and use it to hold the bearing while the loctite sets, the slide it off the other end
the other bearing wouldn't be able to use that tool though
but only one is critical, let the other float
ok guys i left by now, i see later, thansk you by the support!!!
use the dremel solution - cut a slot in it
see you juanjo
jepler: you might just add the 201/202 to the Gecko selection
juanjo is now known as Juanjo
jepler: is there a version number in the stepconf-generated files?
BigJohnT: yeah that is more doable
(so an updated stepconf could recognize an "old style" file)
the "quick-pick" selection is currently done with an index?
SWPLinux: originally I just assumed I'd get the software perfect before anybody else used it
a likely story :)
right, 0 is first entry, and so on
and I made "other" the last entry
hmm a database backend
I had the same problem with signal names on port setup, so for 2.3 I've used strings instead. but I won't make that big a change during 2.2..
if it's changed so a named index is used, it's easy to recognize old-style files because they'll have a number
the structure that contains the timings can also be hacked to have the index of that entry in the last version that used indexes
so an old-style file could still be output if necessary
(and the named entry could be found)
er, for output? nah
ok, you've thought of it all already - I'll shut up now :)
you have a two-way lookup once the entries have both a number and name (duh) :)
anybody know offhand if the 4000/500/20000/1000 timings are OK for geckodrive 211 and 212?
Jebus jimminy cricket... finally.... http://farm3.static.flickr.com/2334/2204778154_06e09f2d96_b.jpg
SWPLinux: Better? ;)
not the best machining on the end of that
looks like crap
it also looks like the turned part is longer than the bearing is thick, so you can't clamp down on the bearing
washer to the screw?
yeah. something like that
I'd probably put it between the bearing and nut, unless its a precision washer
I was able to sneak in that photo befor ethe battery died completely, it's charging now, then I cna take more.
look for lack of shoulder where the bearing contacts (upper in pic) bearring could "lean" eg swash plate motion
needs a damn good fit to stop that with such a thin screw and large bearing
would attempt to bend the screw if bearing is tight in the housing
especially when you pull the two ends toward the screw with nuts *to hold the screw in place)
SWPLinux: Not to mention their "FINE Precision machining" as you saw examples of yesterday.
* Jymm keeps looking over at the charger every 30s waiting for the light to turn green - getting sore neck now :(
it would probably be faster for me to drive over there with my camera
What was that ballscrew cut with ?
could be rolled
probably a hacksaw ;)
looks rolled, see the groove between
(for the turning, and an axe to actually cut it to length)
Oh not the screw but the end ..
like I said, a hacksaw, which must have been hard to mount as a lathe tool ;)
the sad part is that somebody charged money for that
Yeah thats what I was asking about whether the material or the method... jeez
it would be not so bad if it was a homebrew job
OMG that was as supplied?
it's too bad you need to know more to be able to evaluate what someone else has done than to do it ourself
that's my understanding
yeah - it looks like k2cnc isn't the best supplier out there
or at least they weren't a couple of years ago
k2cnc needs a machinist
and an engineer
what is the meaning of the first dotin command . ./scripts/emc-environment
what is the meaning of the first dot in command . ./scripts/emc-environment
it's equiva;lent to the "source" command
umm, then I'll learn what is source
it runs the script in the same environment (or something like that)
yes - otherwise, when the script ended, the environment variables it had created would go away.
acemi: it adds some env variables to the current shell
sources aka learns them
from then on things like 'man comp' will work
I see now, thanks SWPLinux jtr alex_joni
gots a problem booting into the livecd from my other pc.
gets to the 2nd screen after u load kernel. and it goes threw the first action and it says ok then the next and it freezes
does the boot screen say something like 'all actions recorded to blah.log'?
i dont believe so.
i only get to loading cd, then unloadin kernel then the 2nd screen that has color, then it freezes
disregard the 'log' this is live cd, logs vanish
tried switching out motherboards,
differant video cards
sounds vaguely like an x problem, like it didnt like you vid, but i'm no expert
same thing happeneds
differnt mobos AND diff vid cards? sounds like bad cd/cdreader
it does shut down after a while...
u can hear it spin slower and slower.. in succesion
i think you were doing 2 systems so try swapping the cd device to the one that did work
(becuz , if same cd, then its the reader maybe)
cant lol my cd rom on my pc now is sata.
i know the sounds
was old ide and new is ONLY sata?
what a bummer
Is SATA supported in that kernel?
i guess.... it did work....
on this pc atleast
did work, doesnt now? "hardware problem"
or do you mean it never worked with live cd and always works some other way?
live cd worked on my new computer, not on the old one i have for the machine, it does but only goes so far
* maddash is the One b/c he managed to compute the moving avg. and RMS for 3 separate analog channels in <500 clocks on a 29.49MHz Rabbit
try the 'cehck cd' option on the boot screen to make sure the cd reader is good
I have two older generic cdrom drives around the house linux just won't boot on , but the rest of the pc is fine with a newer cd drive
* maddash has a .4+/-.2 ms reaction time
which is freakin' ironic
crap, wrong channel